Identification of seven novel loci associated with amino acid levels using single-variant and gene-based tests in 8545 Finnish men from the METSIM study

Abstract

Comprehensive metabolite profiling captures many highly heritable traits, including amino acid levels, which are potentially sensitive biomarkers for disease pathogenesis. To better understand the contribution of genetic variation to amino acid levels, we performed single variant and gene-based tests of association between nine serum amino acids (alanine, glutamine, glycine, histidine, isoleucine, leucine, phenylalanine, tyrosine, and valine) and 16.6 million genotyped and imputed variants in 8545 non-diabetic Finnish men from the METabolic Syndrome In Men (METSIM) study with replication in Northern Finland Birth Cohort (NFBC1966). We identified five novel loci associated with amino acid levels (P = < 5×10-8): LOC157273/PPP1R3B with glycine (rs9987289, P = 2.3×10-26); ZFHX3 (chr16:73326579, minor allele frequency (MAF) = 0.42%, P = 3.6×10-9), LIPC (rs10468017, P = 1.5×10-8), and WWOX (rs9937914, P = 3.8×10-8) with alanine; and TRIB1 with tyrosine (rs28601761, P = 8×10-9). Gene-based tests identified two novel genes harboring missense variants of MAF <1% that show aggregate association with amino acid levels: PYCR1 with glycine (Pgene = 1.5×10-6) and BCAT2 with valine (Pgene = 7.4×10-7); neither gene was implicated by single variant association tests. These findings are among the first applications of gene-based tests to identify new loci for amino acid levels. In addition to the seven novel gene associations, we identified five independent signals at established amino acid loci, including two rare variant signals at GLDC (rs138640017, MAF=0.95%, Pconditional = 5.8×10-40) with glycine levels and HAL (rs141635447, MAF = 0.46%, Pconditional = 9.4×10-11) with histidine levels. Examination of all single variant association results in our data revealed a strong inverse relationship between effect size and MAF (Ptrend<0.001). These novel signals provide further insight into the molecular mechanisms of amino acid metabolism and potentially, their perturbations in disease.

Figures

Figure 1.

Relationship between minor allele frequency and estimated beta coefficient (β) for loci associated with amino acid levels in the METSIM data. All novel amino acid loci (triangles in pink) are highlighted, in addition to novel signals at known amino acid loci (squares in green) identified through analyses conditioned on all known amino acid GWAS variants. The known amino acid signals are represented with blue circles. The dashed gray line represents a fractional polynomial spline fitted to the data points (P < 0.001). β, on the y-axis, is the absolute value of the estimated regression coefficient for a given variant-trait association.

Figure 2.

Plots show the trait values of rare variant carriers relative to the distribution of amino acid levels in all individuals. The tables in the right panel show gene-based tests of association with amino acid levels for genes PYCR1 and BCAT2. Histograms show the distribution of the inverse normalized residuals of the trait across all participants for the gene-based test of association at (A) PYCR1 with glycine levels and (B) BCAT2 with valine. The dashed gray line represents the mean inverse normalized residual of trait level for all individuals. The solid black line in each row represents the mean trait level for carriers of each variant. Triangles represent rare variant carriers. The locations of triangles relative to the distribution across all participants indicate the trait levels of rare variant carriers. No individuals were homozygous for the minor allele of any of the listed variants.